RAINFALL THRESHOLDS FOR THE OCCURRENCE OF URBAN LANDSLIDES IN TEGUCIGALPA, HONDURAS: AN APPLICATION OF THE CRITICAL RAINFALL INTENSITY

The occurrence of landslides on 134 days between 1980 and 2005 in Tegucigalpa, Honduras has been analyzed in order to determine the contribution of triggering and antecedent rainfall. For each of the 6344 days in the study period, the antecedent rainfall for 19 durations ranging between 1 and 60 days has been determined and the corresponding return period has been calculated using the Gumbel extreme value distribution. The critical rainfall intensity for every landslide day, denoting the antecedent rainfall amount yielding the highest return period divided by its antecedent duration, has been used for the construction of four rainfall thresholds with durations of 7, 15, 30 and 60 days. The accuracy of each threshold in discriminating between landside and non-landslide days is determined by comparing the antecedent rainfall amounts of each day in the study period with the threshold's rainfall requirements. When threshold lines are allowed to underlie 15%, 50% and 85% of all landslide days, the best performance in the receiver operating characteristic space is achieved with the 7-day threshold. To improve the threshold performance, landslide days whose triggering rainfall has higher return periods than their corresponding antecedent rainfall amounts are represented in the line x=0.1 of the critical rainfall intensity-antecedent duration plot. Furthermore, a critical region where minimum triggering and antecedent rainfall amounts are necessary for landslide occurrence has been established to reduce the number of false alarms. Nonetheless, the threshold's low rainfall requirements suggest that human intervention has significantly predisposed Tegucigalpa's slopes to failure.